ForvelSee it work

Eight areas, measuredone named drivera dated prediction

Something
is in the
way.

We write down what. Then we spend months trying to prove ourselves wrong.

See it work

$290 a yearno free tierno card to look

Fig. 1MEASURED
HealthMindWorkMoneyCloseFriendsGrowthPlay

Eight areas, measured Tue 14 Jul. The hatched band is the gap between what you said matters and where it actually is. Two are being worked. The other six are measured, then left alone.

No streaksNo pointsNo levelsNo badgesNo day-countsNo completion %No leaderboardNo score of you

01 It names one thing

Avoidance of a finished thing

Not a type, not a diagnosis and not a history. One named thing, written in a sentence specific enough that a week can contradict it.

v4HYPOTHESIS

You went flat after hitting a goal, and have avoided a twenty-minute task ever since. Not a trait, and not a history — a specific avoidance, of a specific kind of task, that starts the week something closes. The observable part is the twenty minutes: it is the smallest task on the list and it is the one that does not get done.

Strength
strong
Confidence
3 of 5

02 It holds that against the alternatives

Three others fit the same week.

Analysis of Competing Hypotheses, from Heuer. Tested experimentally once, on fifty analysts: mixed evidence it reduces confirmation bias. That is what we know about our own instrument.

Bad week — worse after Sunday.

Consistent with all four. It separated none of them. It is not evidence.

Consistency is cheap — a story that fits is not a story that is true. Only a contradiction carries information, so only contradictions are listed below.

  1. Avoidance of a finished thingnothing contradicts it

  2. Avoidance of a significant other's judgmentcontradicted once

    • MEASUREDFlatness did not track contact, across eleven weeks of a fixed battery.
  3. Sleep timing driftcontradicted twice

    • OBSERVEDThe task avoided is the shortest one on the list, and only after something closes.
    • MEASUREDMCTQ drift held under 20 minutes across the same fourteen nights.
  4. The work is not actually wantedcontradicted twice

    • MEASUREDOutput rose 40% in the peak window while discipline was held constant.
    • OBSERVEDThe task avoided is the shortest one on the list, and only after something closes.
  5. None of thesepredicts nothing, so nothing can contradict it

Standing — the only driver nothing contradicts. Not the same as being right.

03 It writes down what would prove it wrong

A prediction, with a date on it.

Sealed before the fact. It opens itself inside the review — nothing to tap, and no window in which it can be quietly rewritten to match what happened.

SEALEDAvoidance of a finished thingopens Sun 19 Jul, inside the scheduled review — it opens itself

What it is predicting about

MEASURED Peak window 09:20–12:40

IF 09:30 THEN twenty minutes on the algo.

Clock cue — a time you have to notice. The window passed 6 times last week, and this never fired.

Also holding

BoundaryWorkIF 12:40 THEN stop. Mid-sentence if that is where you are.

CopingWorkIF the algo stalls THEN write down where it stalled, and stop.

MethodWhat does the work is the cue’s accessibility, not resolve — which is why the kind of cue is named on every rule here, and why the plan is one thing rather than a list.

04 It marks itself wrong

71%

When we propose this driver, we are wrong 71% of the time. We still propose it, because when it holds it holds hard.

How often we are wrong

Template 17 — “Avoidance of a significant other's judgmentProposed 412 · survived 118 · falsified 294 · median 24 days to falsification. We are wrong 71% of the time we propose this one, and propose it anyway: when it holds, it holds hard.

MethodThe driver list above is Analysis of Competing Hypotheses (Heuer). Tested experimentally once, on fifty analysts: mixed evidence it reduces confirmation bias, and it may increase judgment inconsistency — Dhami et al., 2019. It earns its place by making non-diagnostic evidence visible, which is arithmetic, not by debiasing.

What we were wrong about

  1. v1

    Low discipline

    Falsified Sun 3 May Output rose 40% in the peak window while discipline was held constant.

  2. v2

    Avoidance of a parent's judgment

    Falsified Thu 11 Jun Predicted flatness to track contact. It didn't.

  3. v3

    Under-stimulation in the afternoon block

    Lost to follow-up — Never tested. We stopped asking.

05 What it will not do

Not hidden. Not expressible.

A pre-commit hook fails the build on the vocabulary, so none of the things beside this can be added later by someone with a growth target.

SCHEDULED The washout

The 07:30 implementation intention

Out Mon 20 JulMon 3 Aug. If the behaviour holds without it, the scaffold is retired — finished, not failed. No app in this category subtracts.

  • No streaks

    A run length turns one lapse into a relapse.

  • No points, levels or badges

    There is nothing here to win.

  • No day-counts

    A missed week is data, never a debt.

  • No completion percentage

    You are not a progress bar.

  • No composite score of you

    Two claims about one driver are never collapsed into one number.

  • No leaderboard

    The model is built from your data, about you, and shown to you.

A missed week is a re-entry ramp and data — never a reset, and never a shame trigger.

06 What it costs

A tenth of a coach.

$290 a year, billed once. No monthly plan and no free tier — a free tier makes your attention the thing we have to win, and everything beside this is what winning attention looks like.

One click to cancel, in the app. No retention interstitial, no offer to stay, and nobody to talk to first.

What the year is

Weekly
The check-in. Goal, reality, options, will — and one commitment out of it.
Monthly
Tactics only. The focus does not move here, deliberately.
Quarterly
The re-measure. Wheel re-scored, driver re-formulated, focus rotated.
Annual
Keep, throw, add. Then one direction. Not resolutions.
  • BetterUp, four sessions a month$223/mono published price
  • BetterUp, one session a month$71/mono published price
  • Forvel$290/yearone price, billed once
  • Mindvaults Studio€19/mometered by the message
  • Rocky Pro$10/mofunnel for a B2B platform
  • Noom, annual term$17/moprepay, then renew
See it work first

Nothing to enterthe whole instrument, with real data in it

A model that survives every result is not a model. It is a story. This one has a date on it.

See it work

$290 a yearno free tierno card to look